A feedback control method and system for semiconductor polishing

Abstract

The application provides a kind of semiconductor polishing feedback control method and system, it is related to semiconductor polishing processing technical field, the method comprises: constructing intelligent feedback device, including dot array evolution unit, feedback decision unit and fuzzy analysis unit;Combination dot array evolution unit determines target check dot array;Polishing control is carried out based on collaborative control scheme, synchronous execution is based on target check dot array Sensing monitoring, determine monitoring feedback data;Monitoring feedback data is extracted and fuzzy feedback decision with index matrix group, determine feedback control strategy;Receive subjective feedback instruction, execute feedback polishing regulation and control in conjunction with feedback control strategy.The application can solve the technical problem of poor semiconductor polishing quality due to large amount of monitoring data processing, inability to achieve targeted monitoring, resulting in poor timeliness and accuracy of monitoring feedback, can improve the pertinence and purpose of semiconductor polishing monitoring, and achieve the technical effect of improving the quality of semiconductor polishing.

Description

Technical Field

[0001] This application relates to the field of semiconductor polishing technology, and in particular to a feedback control method and system for semiconductor polishing. Background Technology

[0002] Semiconductor polishing is a surface treatment technique for semiconductor materials, designed to improve the smoothness and gloss of their surfaces, thereby laying the foundation for subsequent semiconductor device manufacturing. To ensure the quality and efficiency of polishing, the polishing process needs to be monitored in real time to detect and address any potential anomalies promptly.

[0003] Currently, existing semiconductor polishing methods suffer from poor timeliness and accuracy in monitoring feedback due to the large amount of data processing required during the process, resulting in poor semiconductor polishing quality. Summary of the Invention

[0004] The purpose of this application is to provide a feedback control method and system for semiconductor polishing, in order to solve the technical problem that existing semiconductor polishing methods suffer from poor timeliness and accuracy of monitoring feedback due to the large amount of monitoring data processed during the process, which makes it impossible to achieve targeted monitoring and resulting in poor semiconductor polishing quality.

[0005] In view of the above problems, this application provides a feedback control method and system for semiconductor polishing.

[0006] In a first aspect, this application provides a feedback control method for semiconductor polishing. The method is implemented through a feedback control system for semiconductor polishing, comprising: determining a collaborative control scheme for polishing a target semiconductor using a polishing equipment group, the polishing equipment group including a polishing robot and auxiliary equipment, the collaborative control scheme having a synchronization timestamp identifier; constructing an intelligent feedback unit, the intelligent feedback unit including a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit, the dot matrix evolution unit establishing a communication connection with a sensing equipment group and performing adaptive sensing-assisted management; determining a target verification dot matrix in response to the sensing equipment group, wherein the verification dot matrix has stage variables based on the entire lifecycle of the scheme; synchronously executing sensing monitoring based on the target verification dot matrix while the intelligent central control system performs periodic dynamic polishing control based on the collaborative control scheme, determining monitoring feedback data; extracting an index matrix group and performing fuzzy feedback decision-making on the monitoring feedback data using the intelligent feedback unit, determining a feedback control strategy; connecting to a remote human-machine interface port to receive subjective feedback commands, and, in conjunction with the feedback control strategy, executing feedback polishing regulation of the target semiconductor based on the intelligent central control system.

[0007] Secondly, this application also provides a feedback control system for semiconductor polishing, used to execute a feedback control method for semiconductor polishing as described in the first aspect, wherein the system includes: a collaborative control scheme determination module, which is used to determine a collaborative control scheme for polishing a target semiconductor in conjunction with a polishing equipment group, the polishing equipment group including a polishing robot and auxiliary equipment, and the collaborative control scheme having a synchronization timestamp identifier; an intelligent feedback device construction module, which is used to construct an intelligent feedback device, the intelligent feedback device including a dot matrix evolution unit, a feedback decision unit and a fuzzy analysis unit, the dot matrix evolution unit establishing a communication connection with the sensing equipment group and performing adaptive sensing-assisted management; and a target verification dot matrix determination module, which is used to determine a collaborative control scheme for polishing a target semiconductor in conjunction with a polishing equipment group, the polishing equipment group including a polishing robot and auxiliary equipment, and the collaborative control scheme having a synchronization timestamp identifier; a target verification dot matrix determination module, which is used to determine a target verification dot matrix in conjunction with a polishing equipment group, the target verification dot matrix determination module being used to determine a collaborative control scheme for polishing a target semiconductor in conjunction with a polishing equipment group, the polishing equipment group including a polishing robot and auxiliary equipment, and the collaborative control scheme having a synchronization timestamp identifier; a target verification dot matrix determination module, which is used to determine a target verification dot matrix in conjunction with a target verification dot matrix determination module, the target verification dot matrix determination module being used to determine a target verification dot matrix in conjunction with a polishing equipment group ... The system comprises: a target verification matrix determined by the matrix evolution unit, responding to the sensing device group, wherein the verification matrix has stage variables based on the entire lifecycle of the scheme; a sensing monitoring module, used to synchronously execute sensing monitoring based on the target verification matrix in conjunction with the periodic dynamic polishing control of the intelligent central control system based on the collaborative control scheme, and determine monitoring feedback data; a fuzzy feedback decision module, used to combine with the intelligent feedback device to extract index matrix groups and perform fuzzy feedback decision on the monitoring feedback data, and determine the feedback control strategy; and a feedback polishing control module, used to connect to a remote human-machine interface port, receive subjective feedback commands, and, in conjunction with the feedback control strategy, execute feedback polishing control of the target semiconductor based on the intelligent central control system.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] A collaborative control scheme for polishing a target semiconductor is determined by combining a polishing equipment group, which includes a polishing robot and auxiliary equipment. The collaborative control scheme has a synchronization timestamp identifier. An intelligent feedback unit is constructed, comprising a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. The dot matrix evolution unit establishes a communication connection with the sensing equipment group and performs adaptive sensing-assisted management. A target verification dot matrix is ​​determined by the dot matrix evolution unit in response to the sensing equipment group. The verification dot matrix has staged variables based on the entire lifecycle of the scheme. Along with the periodic dynamic polishing control of the intelligent central control system based on the collaborative control scheme, synchronous sensing monitoring based on the target verification dot matrix is ​​executed to determine monitoring feedback data. Using the intelligent feedback unit, an index matrix group is extracted from the monitoring feedback data, and fuzzy feedback decision-making is performed to determine a feedback control strategy. A remote human-machine interface is connected to receive subjective feedback commands. Based on the feedback control strategy, the intelligent central control system executes feedback polishing regulation of the target semiconductor. In other words, by combining the polishing equipment group and the surface characteristics and quality of the target semiconductor to optimize the polishing scheme, a collaborative control scheme is obtained, which can improve the adaptability and accuracy of the collaborative control scheme setting. Next, an intelligent feedback unit is constructed, including a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. Further, based on the dot matrix evolution unit, a target verification dot matrix is ​​determined, and accompanied by the periodic dynamic polishing control of the intelligent central control system based on the aforementioned collaborative control scheme, the target semiconductor is synchronously monitored based on the target verification dot matrix to obtain monitoring feedback data. Then, through the intelligent feedback unit, the monitoring feedback data is used to extract the index matrix group and perform fuzzy feedback decision-making, outputting a feedback control strategy. In addition, a remote human-machine interface port is connected to receive subjective feedback commands, and based on the subjective feedback commands and the feedback control strategy, feedback polishing control of the target semiconductor is executed. This can improve the targeting and purposefulness of semiconductor polishing monitoring, reduce the amount of monitoring data processing, and thus improve the timeliness and accuracy of monitoring feedback, enabling timely and accurate control of polishing deviations, ultimately achieving the technical effect of improving semiconductor polishing quality.

[0010] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of a feedback control method for semiconductor polishing according to this application;

[0013] Figure 2 This is a flowchart illustrating the process of determining a collaborative control scheme in a feedback control method for semiconductor polishing according to this application.

[0014] Figure 3 This is a schematic diagram of the feedback control system for semiconductor polishing according to this application.

[0015] Explanation of reference numerals in the attached figures:

[0016] The module includes a collaborative control scheme determination module 11, an intelligent feedback device construction module 12, a target verification matrix determination module 13, a sensing and monitoring module 14, a fuzzy feedback decision-making module 15, and a feedback polishing control module 16. Detailed Implementation

[0017] This application provides a feedback control method and system for semiconductor polishing, solving the technical problem that existing semiconductor polishing methods suffer from poor timeliness and accuracy in monitoring feedback due to the large amount of monitoring data processing required during processing, making targeted monitoring impossible and resulting in poor semiconductor polishing quality. The method improves the targeting and purposefulness of semiconductor polishing monitoring, reduces the amount of monitoring data processing, and thus improves the timeliness and accuracy of monitoring feedback. It allows for timely and accurate adjustment of polishing deviations, ultimately improving the technical effect of semiconductor polishing quality.

[0018] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0019] Example 1

[0020] Please see the appendix Figure 1This application provides a feedback control method for semiconductor polishing, applied to the CMP process mode, the method specifically including the following steps:

[0021] Step 1: Determine a collaborative control scheme for polishing the target semiconductor by combining the polishing equipment group, which includes a polishing robot and auxiliary equipment, and the collaborative control scheme has a synchronization timestamp identifier;

[0022] Specifically, CMP, or Chemical Mechanical Polishing, is a key process in semiconductor manufacturing. This process achieves efficient removal of excess material and global nanoscale planarization of the wafer surface through the synergistic effect of chemical etching and mechanical abrasion. During CMP processing, the etchant reacts chemically with the material surface, causing it to dissolve or oxidize and removing surface irregularities. Simultaneously, abrasive particles rub against the material surface, further removing these irregularities and achieving surface planarization.

[0023] First, the target semiconductor and polishing equipment set are acquired. The target semiconductor can be configured according to actual conditions, and its surface characteristics and surface quality are obtained. The polishing equipment set includes a polishing robot and auxiliary equipment. The core polishing equipment of the polishing robot has high automation and intelligence characteristics. Through programming and precise control, the polishing robot can perform polishing operations continuously and stably. The auxiliary equipment includes a liquid sprayer, cooling equipment, etc. Next, based on the surface characteristics and surface quality of the target semiconductor and the polishing equipment set, a polishing control scheme is optimized to obtain a collaborative control scheme. The collaborative control scheme has a synchronization timestamp identifier, which includes collaborative interaction nodes and a response delay tolerance time zone based on node accuracy.

[0024] By combining the polishing equipment group with the surface characteristics and surface quality of the target semiconductor to optimize the polishing scheme, a collaborative control scheme can be obtained. This can improve the adaptability and accuracy of the collaborative control scheme settings, thereby improving the precision of the collaborative processing control of the target semiconductor.

[0025] Step 2: Construct an intelligent feedback device, which includes a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. The dot matrix evolution unit establishes a communication connection with the sensing device group and performs adaptive sensing-assisted management.

[0026] Specifically, an intelligent feedback device is constructed to monitor the polishing process of the target semiconductor in real time and provide timely feedback and adjustment for abnormal polishing deviations during the monitoring process. The intelligent feedback device includes a lattice evolution unit, a feedback decision unit, and a fuzzy analysis unit. The lattice evolution unit optimizes the polishing monitoring nodes of the target semiconductor; the feedback decision unit locates abnormal polishing; and the fuzzy analysis unit performs compensation analysis and fuzzy feedback decision-making based on abnormal polishing deviations. The lattice evolution unit establishes a communication connection with a sensor group to perform adaptive sensing-assisted management. The sensor group includes multiple monitoring sensors for real-time monitoring of the target semiconductor polishing process. By constructing the intelligent feedback device, support is provided for subsequent real-time monitoring and feedback decision-making in the polishing of the target semiconductor.

[0027] Step 3: Determine the target verification matrix by combining the matrix evolution unit, in response to the sensing device group, wherein the verification matrix has phased variables based on the entire life cycle of the scheme;

[0028] Specifically, a target verification matrix is ​​determined based on the matrix evolution unit. This target verification matrix includes verification timestamps and phased verification indicators. Then, the target verification matrix is ​​communicatively connected to the sensing device group. The verification matrix contains phased variables based on the entire lifecycle of the polishing scheme; that is, the verification matrix includes verification timestamps and phased verification indicators based on the entire polishing scheme lifecycle, and the verification timestamps and phased verification indicators correspond one-to-one. Determining the target verification matrix provides a basis for the next step of synchronous sensing and monitoring throughout the entire target semiconductor polishing process.

[0029] Step 4: Along with the periodic dynamic polishing control of the intelligent central control system based on the aforementioned collaborative control scheme, synchronously execute the sensing and monitoring based on the target verification matrix to determine the monitoring feedback data;

[0030] Specifically, the intelligent central control system is a control system integrating advanced technologies and intelligent algorithms to achieve comprehensive monitoring and precise control of the semiconductor polishing process. This system integrates various sensors, actuators, data processing units, and communication interfaces to achieve automated control, data acquisition and processing, process optimization, and fault diagnosis of the polishing equipment. Using the intelligent central control system, the target semiconductor is subjected to periodic dynamic polishing control based on the aforementioned collaborative control scheme. The entire lifecycle dynamic polishing of the target semiconductor is synchronously monitored through the target verification matrix; that is, at each verification time stamp, the monitoring data is compared for deviation based on stage verification indicators to obtain monitoring feedback data.

[0031] Step 5: Combining the intelligent feedback device, extract the indicator matrix group and perform fuzzy feedback decision-making on the monitoring feedback data to determine the feedback control strategy;

[0032] Specifically, the monitoring feedback data is received by the intelligent feedback device, and the monitoring feedback data is extracted and fuzzy feedback decision is performed by the feedback decision unit and the fuzzy analysis unit. The indicator matrix is ​​obtained by extracting the monitoring feedback data based on the phased verification indicators. Then, fuzzy feedback decision is performed based on the indicator matrix to obtain a feedback control strategy. The feedback control strategy is an optimized control scheme for abnormal polishing processing data of the indicator matrix.

[0033] By extracting sensor monitoring data based on the target verification matrix, the targeting and purposefulness of abnormal polishing monitoring can be improved, while reducing the amount of data processing and the redundancy of data analysis, thereby improving the purposefulness, accuracy and efficiency of feedback control.

[0034] Step Six: Connect to the remote human-machine interface port, receive subjective feedback commands, and, in conjunction with the feedback control strategy, execute the feedback polishing control of the target semiconductor based on the intelligent central control system.

[0035] Specifically, a remote human-machine interface (HMI) port is connected to link the local device with a remote user interface for remote control and interaction. Subjective feedback commands are received through the remote HMI port. These subjective feedback commands refer to a subjective control scheme derived from polishing deviation analysis. Next, a comprehensive control analysis is performed based on the subjective feedback commands and the feedback control strategy. For example, the reliability of the subjective feedback commands and the feedback control strategy is determined through analysis. Reliability characterizes the control accuracy; higher accuracy equates to higher reliability. Then, weighting coefficients are set based on the reliability, and the subjective feedback commands and the feedback control strategy are weighted and calculated to obtain a comprehensive control scheme. Finally, the comprehensive control scheme is fed back to the intelligent central control system, and based on the intelligent central control system, feedback polishing control of the target semiconductor is executed according to the comprehensive control scheme.

[0036] The aforementioned feedback control method for semiconductor polishing is applied to a feedback control system for semiconductor polishing. It can solve the technical problem that existing semiconductor polishing processing methods suffer from poor timeliness and accuracy of monitoring feedback due to the large amount of monitoring data processed during the processing, which makes it impossible to achieve targeted monitoring and resulting in poor semiconductor polishing quality. First, a collaborative control scheme for polishing the target semiconductor is determined by combining the polishing equipment group, which includes a polishing robot and auxiliary equipment. The collaborative control scheme has a synchronization timestamp identifier. Then, an intelligent feedback unit is constructed, comprising a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. The dot matrix evolution unit establishes a communication connection with the sensing equipment group to perform adaptive sensing-assisted management. Next, a target verification dot matrix is ​​determined based on the dot matrix evolution unit, responding to the sensing equipment group. The verification dot matrix contains stage variables based on the entire lifecycle of the scheme. Following this, the intelligent central control system performs periodic dynamic polishing control based on the collaborative control scheme, simultaneously executing sensor monitoring based on the target verification dot matrix to determine monitoring feedback data. Furthermore, using the intelligent feedback unit, the monitoring feedback data is used to extract an index matrix group and perform fuzzy feedback decision-making to determine a feedback control strategy. Finally, a remote human-machine interface is connected to receive subjective feedback commands. Based on the feedback control strategy, the intelligent central control system executes feedback polishing regulation of the target semiconductor. By combining the polishing equipment group and the surface characteristics and quality of the target semiconductor to optimize the polishing scheme, a collaborative control scheme is obtained, which can improve the adaptability and accuracy of the collaborative control scheme setting. Next, an intelligent feedback unit is constructed, including a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. Further, based on the dot matrix evolution unit, a target verification dot matrix is ​​determined, and accompanied by the periodic dynamic polishing control of the intelligent central control system based on the aforementioned collaborative control scheme, the target semiconductor is synchronously monitored based on the target verification dot matrix to obtain monitoring feedback data. Then, through the intelligent feedback unit, the monitoring feedback data is used to extract the index matrix group and perform fuzzy feedback decision-making, outputting a feedback control strategy. In addition, a remote human-machine interface port is connected to receive subjective feedback commands, and based on the subjective feedback commands and the feedback control strategy, feedback polishing regulation of the target semiconductor is executed. This can improve the targeting and purposefulness of semiconductor polishing monitoring, reduce the amount of monitoring data processing, and thus improve the timeliness and accuracy of monitoring feedback, enabling timely and accurate regulation of polishing deviations, ultimately achieving the technical effect of improving semiconductor polishing quality.

[0037] Further details are attached. Figure 2 As shown, step one of this application includes:

[0038] Read the single polishing scheme based on the polishing task, wherein the polishing task is determined based on the surface characteristics and surface quality of the target semiconductor, and the single polishing scheme corresponds one-to-one with the polishing equipment group;

[0039] The interaction nodes of the single polishing scheme are determined and the synchronization timestamps are identified, wherein the synchronization timestamps have a response delay tolerance time zone based on the node accuracy;

[0040] By combining the aforementioned single polishing scheme, and taking collaborative interaction, scheme simplification, and quality optimization as target dimensions, multi-objective optimization is performed to determine the collaborative control scheme. The optimization rule is the good-direction iteration and bad-direction backtracking of random objectives, and the random objective is at least one objective index.

[0041] Specifically, firstly, a single polishing scheme based on a polishing task is read, wherein the polishing task is determined based on the surface characteristics and surface quality of the target semiconductor. The surface characteristics of the semiconductor include properties such as the binding energy of surface states, surface state density, and surface state charge; the surface quality includes parameters such as surface roughness, crystal structure, and surface cleanliness. Expected polishing indices are obtained, and then processing deviation analysis is performed based on the expected polishing indices and the surface characteristics and surface quality of the target semiconductor to generate a target polishing task. Furthermore, the target polishing task is divided according to the polishing equipment group to obtain multiple single polishing schemes, wherein each single polishing scheme corresponds one-to-one with the polishing equipment group.

[0042] The interaction nodes of the multiple individual polishing schemes are identified, where each interaction node refers to a time point of interaction between the schemes. Based on the synchronization timestamp of the interaction node identifier, the node precision of the interaction node is further obtained. The node precision can be set according to the importance of the interaction node, where the higher the importance, the greater the node precision requirement. Then, a response delay tolerance time zone is set according to the node precision. The response delay tolerance time zone represents the allowable error range of response delay. The response delay tolerance time zone is inversely proportional to the node precision, that is, the higher the node precision requirement, the smaller the corresponding response delay tolerance time zone. Finally, the response delay tolerance time zone is embedded in the corresponding synchronization timestamp.

[0043] Based on the aforementioned single polishing scheme, a multi-objective optimization of the polishing scheme is performed with collaborative interaction, scheme simplification, and quality optimization as target dimensions. The collaborative interaction refers to smooth communication and cooperation between the polishing equipment group and the intelligent central control system, as well as the same interaction quality among the polishing equipment groups. The scheme simplification aims to simplify the polishing scheme and reduce its complexity and operational difficulty. The quality optimization aims to improve the polishing quality of the target semiconductor. Then, a target fitness function is constructed based on the collaborative interaction, scheme simplification, and quality optimization directions. In the target fitness function, the weight of each indicator is different and can be set according to actual needs. For example, when the polishing quality requirement is large, the weight of the quality optimization direction is relatively large.

[0044] Based on the target fitness function, the single polishing scheme is optimized using the optimization rule to output the cooperative control scheme. The optimization rule is a random objective of iterative optimization and backtracking. The optimization rule combines random search, iterative optimization and backtracking mechanism to find the optimal solution to the problem. This strategy is usually used to solve complex problems, where the objective function may have multiple local optima, or the search space is large and difficult to traverse directly. First, a target is randomly selected. The random target can be at least one objective indicator. Randomly selecting a target means choosing one or more target points as the starting point at the beginning of the search process. This helps avoid getting trapped in local optima and increases the possibility of finding the global optimum. Next, preferential iteration and negative backoff are performed. Preferential iteration means that once the current target point is determined, the algorithm will try to search in the vicinity of that point to find a better solution. This usually involves applying some iterative rule or optimization algorithm (such as gradient descent, genetic algorithm, etc.) based on the current point to gradually approach a better solution. The iteration process continues until a certain stopping condition is reached (such as reaching a preset number of iterations, the improvement in the quality of the solution is less than a certain threshold, etc.). Negative backoff means that if the quality of the current solution deteriorates during the iteration process (i.e., it is worse than the previous solution), the algorithm will adopt a backoff strategy. Backoff may mean returning to a previous better solution and reselecting the search direction. This helps avoid getting trapped in local optima and allows the algorithm to reposition itself in the search space to find a better solution. This optimization rule combines randomness, iterative optimization, and a backoff mechanism, enabling the algorithm to strike a balance between exploration and utilization. It can both use known information to approximate the optimal solution and avoid getting trapped in local optima through randomness and backoff. Therefore, this strategy has certain advantages in solving complex optimization problems.

[0045] By combining the polishing equipment group with the surface characteristics and surface quality of the target semiconductor to optimize the polishing scheme, a collaborative control scheme can be obtained. This can improve the adaptability and accuracy of the collaborative control scheme settings, thereby improving the precision of the collaborative processing control of the target semiconductor.

[0046] Furthermore, step two of this application includes:

[0047] Identify the surface characteristics of the target semiconductor and determine an initial verification matrix, wherein the initial verification matrix has a first number of matrix points based on a first interval spacing.

[0048] The principle of lattice trend change of the verification lattice is determined, in which the lattice distribution is adjusted by staged polishing elements;

[0049] Based on the initial verification matrix and the matrix trend principle, supervised training of the matrix evolution unit is performed.

[0050] Specifically, the intelligent feedback unit includes a dot matrix evolution unit. First, it identifies and analyzes the surface characteristics of the target semiconductor. This includes observing the external morphological features of the semiconductor, such as particle size, shape, and surface flatness, as well as checking its surface structure, potential defects, or impurities. These surface characteristics have a significant impact on the performance and stability of semiconductor devices. For example, particle size and distribution uniformity affect the electrical and optical properties of the material, while surface defects may hinder electron-hole recombination, affecting device efficiency. Next, an initial verification dot matrix is ​​constructed based on the identification results. By rationally arranging the initial verification dot matrix, it can be ensured that the polishing equipment can accurately position and polish the semiconductor during processing. The initial verification dot matrix has a first dot matrix number based on a first interval spacing, that is, there is a verification distance between adjacent verification points. The initial verification dot matrix includes multiple verification points.

[0051] Further, the dot matrix trend principle for the verification dot matrix is ​​determined. This principle involves adjusting the dot matrix distribution based on staged polishing elements. These staged polishing elements refer to the key verification indicators corresponding to each verification point during the target semiconductor polishing process, which can be set by those skilled in the art according to actual conditions. Then, the initial verification dot matrix is ​​optimized and adjusted according to the dot matrix trend principle. Specifically, by observing the changing trends of the verification point distribution and number throughout the entire cycle, a verification dot matrix suitable for the stage scenario is determined while balancing redundancy and effectiveness. Optimizing and adjusting the initial verification dot matrix by setting the dot matrix trend principle can improve the precision and accuracy of the target verification dot matrix setting, thereby providing a basis for targeted monitoring of semiconductor polishing processes.

[0052] Furthermore, step two of this application also includes:

[0053] The main structure of the fuzzy analysis unit is determined, including a bidirectional fuzzy processing port and a fuzzy inference node;

[0054] The bidirectional fuzzy processing port includes a fuzzification direction and a defuzzification direction. The fuzzy inference node interacts bidirectionally with the bidirectional fuzzy processing port to form a fuzzy analysis loop. The fuzzy inference node is connected to a reasoning knowledge base.

[0055] Retrieve polishing records of the target semiconductor homology, filter and feedback control, determine basic data features and inference decision rules, and use them as the inference knowledge base;

[0056] Based on the polishing records, random samples are selected, and supervised training based on the main structure is performed to determine the fuzzy analysis unit.

[0057] Specifically, the intelligent feedback device includes a fuzzy analysis unit. First, the main structure of the fuzzy analysis unit is determined, which includes a bidirectional fuzzy processing port and a fuzzy inference node. The bidirectional fuzzy processing port includes a fuzzification direction and a defuzzification direction. First, the fuzzification direction converts precise polishing parameters or instructions into a fuzzy form. This is because, in some cases, due to the complexity of the semiconductor surface or the uncertainty of the polishing equipment, precise parameters may not fully meet the polishing requirements. By fuzzification, these parameters can be transformed into fuzzy values ​​with a certain range or probability distribution to better adapt to changes in the actual polishing process. Fuzzification can be based on fuzzy logic or fuzzy set theory. By defining appropriate fuzzy sets and membership functions, precise polishing parameters can be mapped onto fuzzy sets and assigned different weights or probabilities. In this way, the polishing system can make decisions and adjustments based on these fuzzy values ​​to adapt to different polishing conditions and requirements. The defuzzification function converts the fuzzy polishing results or feedback into a precise form. During the polishing process, due to the influence of various factors, the polishing results may be fuzzy or uncertain. The purpose of defuzzification is to extract useful information from these fuzzy results and transform them into precise values ​​or conclusions. Defuzzification can employ various methods, such as the central mean method and the maximum membership method. These methods determine a most representative precise value based on the membership distribution or other statistical characteristics of the fuzzy results. In this way, the polishing system can perform subsequent operations or adjustments based on the defuzzified results to achieve the expected polishing effect.

[0058] The fuzzy inference node and the bidirectional fuzzy processing port interact bidirectionally to form a fuzzy analysis loop. The fuzzy inference node, as the core component, primarily functions to perform inference and calculation based on input fuzzy information or data using preset fuzzy rules to derive corresponding outputs. These fuzzy rules may involve various parameters, conditions, or relationships in the polishing process, and are encoded in the fuzzy inference node to guide the inference process. The bidirectional fuzzy processing port is responsible for the input and output of fuzzy information. It can convert precise information from external or internal systems into fuzzy information for use by the fuzzy inference node; simultaneously, it can also convert the output of the fuzzy inference node into precise information for interaction with other systems or devices. This bidirectional interactive characteristic allows the fuzzy analysis loop to flexibly handle various complex and uncertain information. The fuzzy inference node is connected to a reasoning knowledge base.

[0059] Further, polishing records of semiconductors with similar properties to the target semiconductor are retrieved, i.e., historical polishing records of other semiconductors with the same properties as the target semiconductor are obtained. Feedback control data is extracted from these polishing records to obtain feedback control data. Then, based on the feedback control data, basic data features and inference decision rules are determined. First, the feedback control data covers real-time monitoring data from multiple stages of the polishing process, including physical parameters such as polishing force, polishing speed, temperature, and humidity, as well as quality indicators such as surface roughness, flatness, and optical performance. Then, the real-time monitoring data is analyzed in depth to extract basic data features. This includes preprocessing operations such as data cleaning, denoising, and normalization to eliminate outliers and noise interference. Then, statistical analysis and machine learning methods are used to extract features and reduce dimensionality, obtaining a dataset that represents the essential characteristics of the polishing process. Further, inference decision rules are formulated based on this dataset. These rules are based on factors such as polishing targets, process requirements, and empirical knowledge, combined with the relationships and patterns between data features. The rules can be expressed as a series of conditional statements or algorithm models to guide parameter adjustments and scheme optimization during the polishing process. Finally, an inference knowledge base is constructed based on the basic data features and inference decision rules.

[0060] Then, based on the polishing records, sample training data is randomly selected, and the main structure is subjected to supervised training based on the sample training data to obtain the fuzzy analysis unit that meets the expected training index. The expected training index is the output accuracy index, which can be set according to actual needs.

[0061] Furthermore, step five of this application includes:

[0062] In conjunction with the dot matrix evolution unit, the synchronization of the monitoring feedback data is checked. If the check fails, a first feedback strategy for the time control dimension is generated, wherein the first feedback strategy is identified by the feedback device and the parameter control characteristics.

[0063] If the verification is successful, the indicator features are extracted based on the target verification point matrix, and the indicator matrix group is integrated to determine the indicator matrix group, which corresponds one-to-one with the target verification point matrix.

[0064] Specifically, the monitoring feedback data undergoes index matrix extraction and fuzzy feedback decision-making. First, in conjunction with the dot matrix evolution unit, the monitoring feedback data is synchronized. That is, at each verification time stamp, the monitoring data is extracted and verified based on the expected processing indicators of the verification time stamp to determine whether the monitoring data meets the expected processing indicators. If the monitoring data does not meet the expected processing indicators, indicating a failed verification, a first feedback strategy for the time control dimension is generated. This first feedback strategy identifies a feedback device and control parameters. The feedback device is an abnormal polishing device, and the control parameters are the control parameters of the abnormal polishing device. If the monitoring data meets the expected processing indicators, indicating a successful verification, then index features are extracted from the monitoring data based on the staged polishing elements of the verification points in the target verification dot matrix to obtain an index matrix group. This index matrix group corresponds one-to-one with the target verification dot matrix.

[0065] Furthermore, this application also includes the following steps:

[0066] A matrix distribution of indicators is established, with the control deviation dimension as the first indicator multivariate group and the polishing quality dimension as the second indicator multivariate group.

[0067] Identify the monitoring feedback data and locate the target verification matrix;

[0068] Based on the distribution of the indicator matrix, the indicator feature values ​​are extracted by traversing the target verification matrix, integrated and mapped to generate the indicator matrix group.

[0069] Specifically, firstly, an indicator matrix distribution is constructed using the control deviation dimension as the first indicator multivariate and the polishing quality dimension as the second indicator multivariate. Next, the monitoring feedback data is identified, and the target verification point matrix is ​​located based on the monitoring feedback data, that is, the abnormal verification points of the target verification point matrix are determined. Finally, in combination with the indicator matrix distribution, indicator feature values ​​are extracted from the target verification point matrix, and the indicator feature extraction results are mapped to the matrix to construct the indicator matrix group.

[0070] By extracting indicator feature values ​​from monitoring feedback data based on the target verification matrix, the targeting and purpose of semiconductor polishing monitoring can be improved. At the same time, dynamic adjustment of the verification matrix can ensure the compatibility between the verification matrix and the polishing stage, thereby reducing the amount of monitoring data analysis and processing and improving the efficiency of monitoring data analysis.

[0071] Based on the feedback decision-making unit, the abnormal polishing is located using the single-point quality and global fluctuation trend as a benchmark. The result is then transferred to the fuzzy analysis unit for adjustment and decision-making to determine the feedback control strategy. The abnormal location features include abnormal points and abnormal features.

[0072] Specifically, in conjunction with the feedback decision unit, abnormal polishing is located based on single-point quality and global fluctuation trend. Single-point quality refers to the polishing quality of a single polishing point or area, which is usually evaluated by measuring indicators such as surface roughness and flatness at that point. Global fluctuation trend refers to the changing trend of quality indicators throughout the polishing process, reflecting the overall stability and consistency of polishing quality. Abnormal location features include abnormal points and abnormal characteristics. Then, the abnormal location features are transferred to the fuzzy analysis unit for adjustment decision-making, and a feedback control strategy is output. The feedback control strategy is an optimized scheme for abnormal polishing location.

[0073] Furthermore, step six of this application includes:

[0074] The self-feedback control module of the polishing robot is determined, wherein the self-feedback control module is connected to the force feedback sensor assembled in the organism;

[0075] Establish a collaborative relationship between the self-feedback control module and the intelligent feedback device.

[0076] Specifically, firstly, a self-feedback control module for the polishing robot is defined. This module automatically adjusts system parameters and control strategies by collecting and analyzing the difference between the system output and the desired target in real time, thereby achieving precise control of the polishing process. The self-feedback control module is connected to a force feedback sensor assembled within the robot. This force feedback sensor is a core component of the self-feedback control module; it can detect the magnitude and direction of the force applied by the polishing robot to the semiconductor surface in real time and transmit this information to the self-feedback control module. This real-time force feedback mechanism allows the polishing robot to flexibly adjust the magnitude and distribution of the applied force according to the real-time processing conditions, adapting to different polishing requirements and conditions. That is, after receiving information from the force feedback sensor, the self-feedback control module performs a series of calculations and analyses, comparing the difference between the real-time detected force and the ideal value based on preset polishing parameters and process requirements, and adjusts the polishing robot's motion trajectory, speed, and force parameters accordingly. This real-time feedback and control mechanism enables the polishing robot to maintain a stable processing state during the polishing process, improving polishing quality and efficiency. Finally, a collaborative association is established between the self-feedback control module and the intelligent feedback device. This collaborative association ensures that the self-feedback control module and the intelligent feedback device can interact and share information in real time during the polishing process, and jointly optimize polishing parameters and strategies.

[0077] In summary, the feedback control method for semiconductor polishing provided in this application has the following technical advantages:

[0078] 1. By combining the polishing equipment group and the surface characteristics and quality of the target semiconductor to optimize the polishing scheme, a collaborative control scheme is obtained, which can improve the adaptability and accuracy of the collaborative control scheme setting. Next, an intelligent feedback unit is constructed, including a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. Further, based on the dot matrix evolution unit, a target verification dot matrix is ​​determined, and accompanied by the periodic dynamic polishing control of the intelligent central control system based on the aforementioned collaborative control scheme, the target semiconductor is synchronously monitored based on the target verification dot matrix to obtain monitoring feedback data. Then, through the intelligent feedback unit, the monitoring feedback data is used to extract the index matrix group and perform fuzzy feedback decision-making, outputting a feedback control strategy. In addition, a remote human-machine interface port is connected to receive subjective feedback commands, and based on the subjective feedback commands and the feedback control strategy, feedback polishing control of the target semiconductor is executed. This can improve the targeting and purposefulness of semiconductor polishing monitoring, reduce the amount of monitoring data processing, and thus improve the timeliness and accuracy of monitoring feedback, enabling timely and accurate control of polishing deviations, ultimately achieving the technical effect of improving semiconductor polishing quality.

[0079] 2. By extracting sensor monitoring data based on the target verification matrix, the targeting and purpose of abnormal polishing monitoring can be improved. At the same time, the amount of data processing can be reduced, the redundancy of data analysis can be decreased, thereby improving the purposefulness, accuracy and efficiency of feedback control.

[0080] Example 2

[0081] Based on the feedback control method for semiconductor polishing described in the foregoing embodiments, and using the same inventive concept, this application also provides a feedback control system for semiconductor polishing. Please refer to the appendix. Figure 3 The system includes:

[0082] The collaborative control scheme determination module 11 is used to determine a collaborative control scheme for polishing the target semiconductor in conjunction with the polishing equipment group. The polishing equipment group includes a polishing robot and auxiliary equipment. The collaborative control scheme has a synchronization timestamp identifier.

[0083] The intelligent feedback device construction module 12 is used to construct an intelligent feedback device. The intelligent feedback device includes a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. The dot matrix evolution unit establishes a communication connection with the sensing device group and performs adaptive sensing-assisted management.

[0084] The target verification matrix determination module 13 is used to determine the target verification matrix in conjunction with the matrix evolution unit, in response to the sensing device group, wherein the verification matrix has stage variables based on the entire life cycle of the scheme;

[0085] The sensing and monitoring module 14 is used to synchronously perform sensing and monitoring based on the target verification matrix, along with the periodic dynamic polishing control of the intelligent central control system based on the collaborative control scheme, and to determine the monitoring feedback data.

[0086] Fuzzy feedback decision module 15 is used to combine the intelligent feedback device to extract the indicator matrix group and perform fuzzy feedback decision on the monitoring feedback data to determine the feedback control strategy.

[0087] The feedback polishing control module 16 is used to connect to a remote human-machine interface port, receive subjective feedback commands, and, in conjunction with the feedback control strategy, execute feedback polishing control of the target semiconductor based on the intelligent central control system.

[0088] Furthermore, the cooperative control scheme determination module 11 in the system is also used for:

[0089] Read the single polishing scheme based on the polishing task, wherein the polishing task is determined based on the surface characteristics and surface quality of the target semiconductor, and the single polishing scheme corresponds one-to-one with the polishing equipment group;

[0090] The interaction nodes of the single polishing scheme are determined and the synchronization timestamps are identified, wherein the synchronization timestamps have a response delay tolerance time zone based on the node accuracy;

[0091] By combining the aforementioned single polishing scheme, and taking collaborative interaction, scheme simplification, and quality optimization as target dimensions, multi-objective optimization is performed to determine the collaborative control scheme. The optimization rule is the good-direction iteration and bad-direction backtracking of random objectives, and the random objective is at least one objective index.

[0092] Furthermore, the intelligent feedback device construction module 12 in the system is also used for:

[0093] Identify the surface characteristics of the target semiconductor and determine an initial verification matrix, wherein the initial verification matrix has a first number of matrix points based on a first interval spacing.

[0094] The principle of lattice trend change of the verification lattice is determined, in which the lattice distribution is adjusted by staged polishing elements;

[0095] Based on the initial verification matrix and the matrix trend principle, supervised training of the matrix evolution unit is performed.

[0096] Furthermore, the intelligent feedback device construction module 12 in the system is also used for:

[0097] The main structure of the fuzzy analysis unit is determined, including a bidirectional fuzzy processing port and a fuzzy inference node;

[0098] The bidirectional fuzzy processing port includes a fuzzification direction and a defuzzification direction. The fuzzy inference node interacts bidirectionally with the bidirectional fuzzy processing port to form a fuzzy analysis loop. The fuzzy inference node is connected to a reasoning knowledge base.

[0099] Retrieve polishing records of the target semiconductor homology, filter and feedback control, determine basic data features and inference decision rules, and use them as the inference knowledge base;

[0100] Based on the polishing records, random samples are selected, and supervised training based on the main structure is performed to determine the fuzzy analysis unit.

[0101] Furthermore, the fuzzy feedback decision module 15 in the system is also used for:

[0102] In conjunction with the dot matrix evolution unit, the synchronization of the monitoring feedback data is checked. If the check fails, a first feedback strategy for the time control dimension is generated, wherein the first feedback strategy is identified by the feedback device and the parameter control characteristics.

[0103] If the verification is successful, the indicator features are extracted based on the target verification point matrix, and the indicator matrix group is integrated to determine the indicator matrix group, which corresponds one-to-one with the target verification point matrix.

[0104] Based on the feedback decision-making unit, the abnormal polishing is located using the single-point quality and global fluctuation trend as a benchmark. The result is then transferred to the fuzzy analysis unit for adjustment and decision-making to determine the feedback control strategy. The abnormal location features include abnormal points and abnormal features.

[0105] Furthermore, the fuzzy feedback decision module 15 in the system is also used for:

[0106] A matrix distribution of indicators is established, with the control deviation dimension as the first indicator multivariate group and the polishing quality dimension as the second indicator multivariate group.

[0107] Identify the monitoring feedback data and locate the target verification matrix;

[0108] Based on the distribution of the indicator matrix, the indicator feature values ​​are extracted by traversing the target verification matrix, integrated and mapped to generate the indicator matrix group.

[0109] Furthermore, the feedback polishing control module 16 in the system is also used for:

[0110] The self-feedback control module of the polishing robot is determined, wherein the self-feedback control module is connected to the force feedback sensor assembled in the organism;

[0111] Establish a collaborative relationship between the self-feedback control module and the intelligent feedback device.

[0112] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. The feedback control method and specific examples for semiconductor polishing described in Embodiment 1 are also applicable to the feedback control system for semiconductor polishing in this embodiment. Through the foregoing detailed description of the feedback control method for semiconductor polishing, those skilled in the art can clearly understand the feedback control system for semiconductor polishing in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here. As for the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and relevant parts can be referred to in the method section.

[0113] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0114] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A feedback control method for semiconductor polishing, characterized in that, Applied to CMP process mode, the method includes: A collaborative control scheme for polishing a target semiconductor by combining a polishing equipment group, wherein the polishing equipment group includes a polishing robot and auxiliary equipment, and the collaborative control scheme has a synchronization timestamp identifier; An intelligent feedback device is constructed, which includes a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. The dot matrix evolution unit establishes a communication connection with the sensing device group and performs adaptive sensing-assisted management. The target verification matrix is ​​determined by combining the matrix evolution unit, in response to the sensing device group, wherein the verification matrix has stage variables based on the entire life cycle of the scheme; Along with the periodic dynamic polishing control of the intelligent central control system based on the aforementioned collaborative control scheme, the sensing and monitoring based on the target verification matrix are executed synchronously to determine the monitoring feedback data; In conjunction with the intelligent feedback device, the monitoring feedback data is used to extract the indicator matrix group and perform fuzzy feedback decision-making to determine the feedback control strategy; Connect to a remote human-machine interface port, receive subjective feedback commands, and, in conjunction with the feedback control strategy, execute feedback polishing control of the target semiconductor based on the intelligent central control system; The intelligent feedback device includes a dot matrix evolution unit, comprising: Identify the surface characteristics of the target semiconductor and determine an initial verification matrix, wherein the initial verification matrix has a first number of matrix points based on a first interval spacing. The principle of lattice trend change of the verification lattice is determined, in which the lattice distribution is adjusted by staged polishing elements; Based on the initial verification matrix and the matrix trend principle, supervised training of the matrix evolution unit is performed; The intelligent feedback device includes a fuzzy analysis unit, comprising: The main structure of the fuzzy analysis unit is determined, including a bidirectional fuzzy processing port and a fuzzy inference node; The bidirectional fuzzy processing port includes a fuzzification direction and a defuzzification direction. The fuzzy inference node interacts bidirectionally with the bidirectional fuzzy processing port to form a fuzzy analysis loop. The fuzzy inference node is connected to a reasoning knowledge base. Retrieve polishing records of the target semiconductor homology, filter and feedback control, determine basic data features and inference decision rules, and use them as the inference knowledge base; Based on the polishing records, random samples are selected, and supervised training based on the main structure is performed to determine the fuzzy analysis unit.

2. The method as described in claim 1, characterized in that, The coordinated control scheme for determining the target semiconductor for polishing includes: Read the single polishing scheme based on the polishing task, wherein the polishing task is determined based on the surface characteristics and surface quality of the target semiconductor, and the single polishing scheme corresponds one-to-one with the polishing equipment group; The interaction nodes of the single polishing scheme are determined and the synchronization timestamps are identified, wherein the synchronization timestamps have a response delay tolerance time zone based on the node accuracy; By combining the aforementioned single polishing scheme, and taking collaborative interaction, scheme simplification, and quality optimization as target dimensions, multi-objective optimization is performed to determine the collaborative control scheme. The optimization rule is the good-direction iteration and bad-direction backtracking of random objectives, and the random objective is at least one objective index.

3. The method as described in claim 1, characterized in that, Extracting the indicator matrix group and performing fuzzy feedback decision-making on the monitoring feedback data includes: In conjunction with the dot matrix evolution unit, the synchronization of the monitoring feedback data is checked. If the check fails, a first feedback strategy for the time control dimension is generated, wherein the first feedback strategy is identified by the feedback device and the parameter control characteristics. If the verification is successful, the indicator features are extracted based on the target verification point matrix, and the indicator matrix group is integrated to determine the indicator matrix group, which corresponds one-to-one with the target verification point matrix. Based on the feedback decision-making unit, the abnormal polishing is located using the single-point quality and global fluctuation trend as a benchmark. The result is then transferred to the fuzzy analysis unit for adjustment and decision-making to determine the feedback control strategy. The abnormal location features include abnormal points and abnormal features.

4. The method as described in claim 3, characterized in that, A matrix distribution of indicators is established, with the control deviation dimension as the first indicator multivariate group and the polishing quality dimension as the second indicator multivariate group. Identify the monitoring feedback data and locate the target verification matrix; Based on the distribution of the indicator matrix, the indicator feature values ​​are extracted by traversing the target verification matrix, integrated and mapped to generate the indicator matrix group.

5. The method as described in claim 1, characterized in that, include: The self-feedback control module of the polishing robot is determined, wherein the self-feedback control module is connected to the force feedback sensor assembled in the organism; Establish a collaborative relationship between the self-feedback control module and the intelligent feedback device.

6. A feedback control system for semiconductor polishing, characterized in that, The system comprises: steps for implementing the method according to any one of claims 1 to 5; A collaborative control scheme determination module is used to determine a collaborative control scheme for polishing a target semiconductor in conjunction with a polishing equipment group. The polishing equipment group includes a polishing robot and auxiliary equipment. The collaborative control scheme has a synchronization timestamp identifier. The intelligent feedback device construction module is used to construct an intelligent feedback device, which includes a dot matrix evolution unit, a feedback decision unit, and a fuzzy analysis unit. The dot matrix evolution unit establishes a communication connection with the sensing device group and performs adaptive sensing-assisted management. A target verification matrix determination module is used to determine the target verification matrix in conjunction with the matrix evolution unit, in response to the sensing device group, wherein the verification matrix has stage variables based on the entire life cycle of the scheme; The sensing and monitoring module is used to synchronously perform sensing and monitoring based on the target verification matrix, along with the periodic dynamic polishing control of the intelligent central control system based on the collaborative control scheme, and to determine the monitoring feedback data. A fuzzy feedback decision module is used to combine the intelligent feedback device to extract the indicator matrix group and perform fuzzy feedback decision on the monitoring feedback data to determine the feedback control strategy. A feedback polishing control module is used to connect to a remote human-machine interface port, receive subjective feedback commands, and, in conjunction with the feedback control strategy, execute feedback polishing control of the target semiconductor based on the intelligent central control system.

Images